Texture feature extraction for ultrasonic images based on improved Gabor-ring / 医疗卫生装备

ABSTRACT

Gabor-ring is often used in texture feature extraction,but it is based on the assumption that texture is narrowband signal and can not applied to most actual textured images that have more than one center frequency in power spectrum.To overcome the limitation of existing method,an improved method is proposed.According to this method,Gabor filters are constructed in accordance with the spectrum feature vector of texture,and a relaxation parameter is used to determine the effect of difference filters so as to integrate the frequencies and phases into the process of extraction.Experimental results of 114 two-dimensional echocardiography images show that the improved algorithm is feasible.